2 /* (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
10 #include <linux/types.h>
12 #include <linux/netfilter.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/icmp.h>
16 #include <linux/sysctl.h>
17 #include <net/route.h>
20 #include <linux/netfilter_ipv4.h>
21 #include <net/netfilter/nf_conntrack.h>
22 #include <net/netfilter/nf_conntrack_helper.h>
23 #include <net/netfilter/nf_conntrack_l4proto.h>
24 #include <net/netfilter/nf_conntrack_l3proto.h>
25 #include <net/netfilter/nf_conntrack_zones.h>
26 #include <net/netfilter/nf_conntrack_core.h>
27 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
28 #include <net/netfilter/nf_nat_helper.h>
29 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
30 #include <net/netfilter/nf_log.h>
32 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
34 enum ip_conntrack_info ctinfo);
35 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
37 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
38 struct nf_conntrack_tuple *tuple)
42 ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
43 sizeof(u_int32_t) * 2, _addrs);
47 tuple->src.u3.ip = ap[0];
48 tuple->dst.u3.ip = ap[1];
53 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
54 const struct nf_conntrack_tuple *orig)
56 tuple->src.u3.ip = orig->dst.u3.ip;
57 tuple->dst.u3.ip = orig->src.u3.ip;
62 static int ipv4_print_tuple(struct seq_file *s,
63 const struct nf_conntrack_tuple *tuple)
65 return seq_printf(s, "src=%pI4 dst=%pI4 ",
66 &tuple->src.u3.ip, &tuple->dst.u3.ip);
69 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
70 unsigned int *dataoff, u_int8_t *protonum)
72 const struct iphdr *iph;
75 iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
79 /* Conntrack defragments packets, we might still see fragments
80 * inside ICMP packets though. */
81 if (iph->frag_off & htons(IP_OFFSET))
84 *dataoff = nhoff + (iph->ihl << 2);
85 *protonum = iph->protocol;
87 /* Check bogus IP headers */
88 if (*dataoff > skb->len) {
89 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
90 "nhoff %u, ihl %u, skblen %u\n",
91 nhoff, iph->ihl << 2, skb->len);
98 static unsigned int ipv4_confirm(unsigned int hooknum,
100 const struct net_device *in,
101 const struct net_device *out,
102 int (*okfn)(struct sk_buff *))
105 enum ip_conntrack_info ctinfo;
106 const struct nf_conn_help *help;
107 const struct nf_conntrack_helper *helper;
110 /* This is where we call the helper: as the packet goes out. */
111 ct = nf_ct_get(skb, &ctinfo);
112 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
115 help = nfct_help(ct);
119 /* rcu_read_lock()ed by nf_hook_slow */
120 helper = rcu_dereference(help->helper);
124 ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
126 if (ret != NF_ACCEPT) {
127 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
128 "nf_ct_%s: dropping packet", helper->name);
132 /* adjust seqs for loopback traffic only in outgoing direction */
133 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
134 !nf_is_loopback_packet(skb)) {
135 typeof(nf_nat_seq_adjust_hook) seq_adjust;
137 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
138 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
139 NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
144 /* We've seen it coming out the other side: confirm it */
145 return nf_conntrack_confirm(skb);
148 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
150 const struct net_device *in,
151 const struct net_device *out,
152 int (*okfn)(struct sk_buff *))
154 return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
157 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
159 const struct net_device *in,
160 const struct net_device *out,
161 int (*okfn)(struct sk_buff *))
163 /* root is playing with raw sockets. */
164 if (skb->len < sizeof(struct iphdr) ||
165 ip_hdrlen(skb) < sizeof(struct iphdr))
167 return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
170 /* Connection tracking may drop packets, but never alters them, so
171 make it the first hook. */
172 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
174 .hook = ipv4_conntrack_in,
175 .owner = THIS_MODULE,
177 .hooknum = NF_INET_PRE_ROUTING,
178 .priority = NF_IP_PRI_CONNTRACK,
181 .hook = ipv4_conntrack_local,
182 .owner = THIS_MODULE,
184 .hooknum = NF_INET_LOCAL_OUT,
185 .priority = NF_IP_PRI_CONNTRACK,
188 .hook = ipv4_confirm,
189 .owner = THIS_MODULE,
191 .hooknum = NF_INET_POST_ROUTING,
192 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
195 .hook = ipv4_confirm,
196 .owner = THIS_MODULE,
198 .hooknum = NF_INET_LOCAL_IN,
199 .priority = NF_IP_PRI_CONNTRACK_CONFIRM,
203 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
204 static int log_invalid_proto_min = 0;
205 static int log_invalid_proto_max = 255;
207 static ctl_table ip_ct_sysctl_table[] = {
209 .procname = "ip_conntrack_max",
210 .data = &nf_conntrack_max,
211 .maxlen = sizeof(int),
213 .proc_handler = proc_dointvec,
216 .procname = "ip_conntrack_count",
217 .data = &init_net.ct.count,
218 .maxlen = sizeof(int),
220 .proc_handler = proc_dointvec,
223 .procname = "ip_conntrack_buckets",
224 .data = &init_net.ct.htable_size,
225 .maxlen = sizeof(unsigned int),
227 .proc_handler = proc_dointvec,
230 .procname = "ip_conntrack_checksum",
231 .data = &init_net.ct.sysctl_checksum,
232 .maxlen = sizeof(int),
234 .proc_handler = proc_dointvec,
237 .procname = "ip_conntrack_log_invalid",
238 .data = &init_net.ct.sysctl_log_invalid,
239 .maxlen = sizeof(unsigned int),
241 .proc_handler = proc_dointvec_minmax,
242 .extra1 = &log_invalid_proto_min,
243 .extra2 = &log_invalid_proto_max,
247 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
249 /* Fast function for those who don't want to parse /proc (and I don't
251 /* Reversing the socket's dst/src point of view gives us the reply
254 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
256 const struct inet_sock *inet = inet_sk(sk);
257 const struct nf_conntrack_tuple_hash *h;
258 struct nf_conntrack_tuple tuple;
260 memset(&tuple, 0, sizeof(tuple));
261 tuple.src.u3.ip = inet->inet_rcv_saddr;
262 tuple.src.u.tcp.port = inet->inet_sport;
263 tuple.dst.u3.ip = inet->inet_daddr;
264 tuple.dst.u.tcp.port = inet->inet_dport;
265 tuple.src.l3num = PF_INET;
266 tuple.dst.protonum = sk->sk_protocol;
268 /* We only do TCP and SCTP at the moment: is there a better way? */
269 if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
270 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
274 if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
275 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
276 *len, sizeof(struct sockaddr_in));
280 h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
282 struct sockaddr_in sin;
283 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
285 sin.sin_family = AF_INET;
286 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
287 .tuple.dst.u.tcp.port;
288 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
290 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
292 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
293 &sin.sin_addr.s_addr, ntohs(sin.sin_port));
295 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
300 pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
301 &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
302 &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
306 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
308 #include <linux/netfilter/nfnetlink.h>
309 #include <linux/netfilter/nfnetlink_conntrack.h>
311 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
312 const struct nf_conntrack_tuple *tuple)
314 NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
315 NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
322 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
323 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
324 [CTA_IP_V4_DST] = { .type = NLA_U32 },
327 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
328 struct nf_conntrack_tuple *t)
330 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
333 t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
334 t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
339 static int ipv4_nlattr_tuple_size(void)
341 return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
345 static struct nf_sockopt_ops so_getorigdst = {
347 .get_optmin = SO_ORIGINAL_DST,
348 .get_optmax = SO_ORIGINAL_DST+1,
350 .owner = THIS_MODULE,
353 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
356 .pkt_to_tuple = ipv4_pkt_to_tuple,
357 .invert_tuple = ipv4_invert_tuple,
358 .print_tuple = ipv4_print_tuple,
359 .get_l4proto = ipv4_get_l4proto,
360 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
361 .tuple_to_nlattr = ipv4_tuple_to_nlattr,
362 .nlattr_tuple_size = ipv4_nlattr_tuple_size,
363 .nlattr_to_tuple = ipv4_nlattr_to_tuple,
364 .nla_policy = ipv4_nla_policy,
366 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
367 .ctl_table_path = nf_net_ipv4_netfilter_sysctl_path,
368 .ctl_table = ip_ct_sysctl_table,
373 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
374 &nf_conntrack_htable_size, 0600);
376 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
377 MODULE_ALIAS("ip_conntrack");
378 MODULE_LICENSE("GPL");
380 static int __init nf_conntrack_l3proto_ipv4_init(void)
385 nf_defrag_ipv4_enable();
387 ret = nf_register_sockopt(&so_getorigdst);
389 printk(KERN_ERR "Unable to register netfilter socket option\n");
393 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
395 pr_err("nf_conntrack_ipv4: can't register tcp.\n");
396 goto cleanup_sockopt;
399 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
401 pr_err("nf_conntrack_ipv4: can't register udp.\n");
405 ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
407 pr_err("nf_conntrack_ipv4: can't register icmp.\n");
411 ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
413 pr_err("nf_conntrack_ipv4: can't register ipv4\n");
417 ret = nf_register_hooks(ipv4_conntrack_ops,
418 ARRAY_SIZE(ipv4_conntrack_ops));
420 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
423 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
424 ret = nf_conntrack_ipv4_compat_init();
429 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
431 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
434 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
436 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
438 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
440 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
442 nf_unregister_sockopt(&so_getorigdst);
446 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
449 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
450 nf_conntrack_ipv4_compat_fini();
452 nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
453 nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
454 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
455 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
456 nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
457 nf_unregister_sockopt(&so_getorigdst);
460 module_init(nf_conntrack_l3proto_ipv4_init);
461 module_exit(nf_conntrack_l3proto_ipv4_fini);
463 void need_ipv4_conntrack(void)
467 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);